FrostyTech typically eschews real world testing for synthetic thermal test platforms; repeatability is more controllable, and parameters like thermal load can be specified from the start by selecting appropriate resistive heater components.

However, when it comes to videocard heatsinks it has proven a challenge to design a suitable platform which will accommodate the wide variety of mounting styles, GPU core dimensions, and VGA heatsinks out there.... Videocard heatsinks react strongly to the internal case environments, and when front/back thermal solutions start to be considered the equation gets rather complicated.

Thus for the time being, FrostyTech will be conducting its VGA heatsink tests with an actual videocard. Temperatures will be recorded with the computer system at idle on the desktop, and when stressed with 3DMark06 where the peak temperature will be noted. The parameters of equipment set up and instrumentation are noted below. The reference heatsink is a stock Albatron FX5900 copper GPU cooler which is pictured at left (no model numbers are available).

FrostyTech.com VGA Heatsink Test Results Results based on readings from a thermocouple positioned on the backside of a FX5900 videocard, directly behind the GPU. The thermocouple is separated from the PCB by Thermagon thermoconductive gapfiller, and insulated on the opposite side by closed cell neoprene. Thermal Test values for each heatsink were calculated by determining the rise over ambient temperature at time of test. Information on Frostytech's test methodology is available here. ATI / nVidia compatible GPU heatsinks
manufacturer:	model no.:	*fan speed:	idle	fx5900 thermal test (°c)	noise level (dba)
Zalman	FC-ZV9	high	3.5	11.5	51.9
Zalman	VF900-Cu	high	3.8	12.0	42.0
Zalman	FC-ZV9	low	8.1	14.5	40.0
Zalman	VF900-Cu	low	8.6	15.2	33.5
Zalman	VF700-AlCu	high	6.4	15.5	43.1
Zalman	VF700-AlCu LED	high	8.9	15.8	43.3
Zalman	VF700-AlCu LED	low	11.9	20.5	32.9
Zalman	VF700-AlCu	low	10.8	21.3	32.6
Albatron	Stock FX5900 GPU HSF		13.6	25.7	48.0
			idle	fx5900 thermal test (°c)	noise level (dba)
*heatsinks are ranked according to the thermal test results column (rise over ambient temp.). low temperatures with low noise levels are considered best. for reference heatsinks with variable-speed fans, only the high speed (12v) fan test result is included in the comparison sheet; more detailed results reside in each specific heatsink review.

there are not a lot of vga heatsinks in this comparison, but what we've tested thus far gives a good indication of where the Zalman FC-ZV9 stands. The FC-ZV9 heatsink posts a 3DMark06 system stressed maximum rise above ambient temperature of 11.5°C, with 51.9 dBA noise (fan speed 3600RPM). With the FC-ZV9's 80mm fan spinning at full speed we see an idle temperature of about 4°C above ambient - idle being when the videocard is just displaying the WindowsXP desktop.

For quieter operation the Zalman Fan Mate 2 is used to reduce the fan speed of the FC-ZV9 to about 1650RPM, thus dropping noise levels to at best 40.0 dBA. In this situation the recorded temperature from the back of the graphics card increases to 14.5°C above ambient under the 3Dmark06 stressed parameter. At idle, the temperature hovers around 8.1°C above ambient.

The stock FX5900 GPU heatsink does no better than 25.7°C above ambient, with 48 dBA noise, under the same conditions.

The Zalman FC-ZV9 videocard heatsink takes a unit approach to GPU cooling, both in materials and thermal design. From our real-world test results it's clear the FC-ZV9 provides better temperature and acoustic results than the stock FX5900 videocard heatsink will. Given the set of data we have to draw conclusions from, I think we can safely say that the Zalman FC-ZV9 heatpipe-based videocard GPU heatsink certainly offers a competitive advantage to GPU cooling.